In GimpProjection's chunk renderer, when the chunk height changes
in the middle of a row, we need to merge the remainder of the
current render area back into the renderer's update region, and
refetch the remainder of the row as the new render area, so that we
don't miss any unrendered area, or re-render already-rendered area,
due to the change in chunk height. However, we should previously
fail to verify that the fetched area is, in fact, the remainder of
the current row, which could cause us to render the wrong area,
missing parts of the update region.
Fix this, by breaking up some of the chunk-renderer fucntions into
smaller sub-functions, and using those in order to explicitly set
the new render area to the remainder of the current row when the
chunk height changes. This also avoids erroneously merging the
unflushed update region of the projection into the renderer's
update region.
(cherry picked from commit c9c2397b0d)
Actually, image grids are saved as parasites, so even though older
GIMP versions round their coordinates upon loading, they maintain
the fractional coordinates when re-saving the image, hence bumping
the XCF version is not really necessary.
This reverts commit 13119efda33a7aba323dc13e6a56207a15a9f000.
(cherry picked from commit 411ddb7e48)
Fractional-coordinate support for image grids was added in commit
1572bccc9f, right before the
introduction of XCF version 10. While images with fractional grid
coordinates can be loaded with earilier versions of GIMP, the grid
coordinates are rounded to the nearest integer.
Bump the minimal XCF version when saving images with fractional
grid coordinates to 10, which should have been the case all along.
(cherry picked from commit a90322278d)
Add a boolean "direct" parameter to gimp_projection_flush_now(),
which specifies if the projection buffer should only be invalidated
(FALSE), or rendered directly (TRUE).
Pass TRUE when flushing the projection during painting, so that the
affected regions are rendered in a single step, instead of tile-by-
tile. We previously only invalidated the projection buffer, but
since we synchronously flush the display right after that, the
invalidated regions would still get rendered, albeit less
efficiently.
Likewise, pass TRUE when benchmarking the projection through the
debug action, and avoid flushing the display, to more accurately
measure the render time.
(cherry picked from commit dac9bfe334)
Don't use a direct-buffer fill if the mode is subtractive, and the
composite region includes the source. Currently, this never
actually happens.
(cherry picked from commit 660f53d300)
In gimp_drawable_edit_fill(), when filling/clearing the whole
drawable, without any special compositing (i.e., when there's no
selection, the opacity is 100%, and the layer mode is trivial),
fill/clear the drawable's buffer directly, without using an
applicator. This makes such operations much faster, especially in
big images.
(cherry picked from commit dd8268c0a2)
... which is similar to gimp_fill_options_create_buffer(), however,
it fills an existing buffer, instead of creating a new buffer.
Implement gimp_fill_options_create_buffer() in terms of the new
function.
(cherry picked from commit 45fc4cb4f9)
When creating a drawable undo from the drawable's buffer, align the
copied rectangle to the buffer's tile grid, so that all the copied
tiles are COWed, saving memory and gaining speed.
Add applied_x and applied_y fields to GimpDrawableUndo, specifying
the position at which to apply the applied_buffer, so that we apply
it in the right place, even if the undo rect has changed due to
alignment.
(cherry picked from commit bb9dd049fb)
Add a scratch-total variable to the dashboard's misc group, showing
the total amount of memory used by the scratch allocator.
(cherry picked from commit 698d1af798)
gimp-scratch is a fast memory allocator (on the order of magnitude
of alloca()), suitable for small (up to a few megabytes), short-
lived (usually, bound to the current stack-frame) allocations.
Unlike alloca(), gimp-scratch doesn't use the stack, and is
therefore safer, and will also serve bigger requests, by falling-
back to malloc().
The allocator itself is very simple: We keep a per-thread stack of
cached memory blocks (allocated using the normal allocator). When
serving an allocation request, we simply pop the top block off the
stack, and return it. If the block is too small, we replace it with
a big-enough block. When the block is freed, we push it back to
the top of the stack (note that even though each thread uses a
separate stack, blocks can be migrated between threads, i.e.,
allocated on one thread, and freed on another thread, although this
is not really an intended usage pattern.) The idea is that the
stacks will ultimately stabalize to contain blocks that can serve
all the encountered allocation patterns, without needing to reisze
any of the blocks; as a consequence, the amount of scratch memory
allocated at any given time should really be kept to a minimum.
(cherry picked from commit a8a8655285)
... which is similar to gimp_async_add_callback(), taking an
additional GObject argument. The object is kept alive for the
duration of the callback, and the callback is automatically removed
when the object is destroyed (if it hasn't been already called).
This is analogous to g_signal_connect_object(), compared to
g_signal_connect().
(cherry picked from commit 49fd2847ac)
In gimp_async_remove_callback(), if removing the last callback
while the callback idle-source is already pending, cancel the idle
source and unref the async object (the async is reffed when adding
the idle source.)
(cherry picked from commit a779dd3849)
Use gimp_tile_handler_validate_validate(), added in the last
commit, in GimpProjection, in order to render the projection,
instead of separately invalidating the buffer, undoing the
invalidation, and then rendering the graph. This is more
efficient, and more idiomatic.
(cherry picked from commit d6f0ca5531)
Add begin_validate() and end_validate() virtual functions, and
corresponding free functions, to GimpTileHandlerValidate. These
functions are called before/after validation happens, and should
perform any necessary steps to prepare for validation. The default
implementation suspends validation on tile access, so that the
assigned buffer may be accessed without causing validation.
Implement the new functions in GimpTileHandlerProjectable, by
calling gimp_projectable_begin_render() and
gimp_projectable_end_render(), respectively, instead of calling
these functions in the ::validate() implementation (which, in turn,
allows us to use the default ::validate() implementation.)
In GimpProjection, use the new functions in place of
gimp_projectable_{begin,end}_render().
(cherry picked from commit 5a623fc54b)
In gimp_projection_finish_draw(), make sure we don't accidentally
re-start the chunk renderer idle source while running the remaining
iterations, in case the chunk height changes, and we need to reinit
the renderer state.
(cherry picked from commit 8a47b68194)
Don't needlessly flush projections whose buffer hasn't been
allocated yet. This can happen when opening an image, in which
case the image is flushed before its projection has a buffer.
(cherry picked from commit b07f810273)
When an async that was created through
gimp_parallel_run_async[_full](), and whose execution is still
pending, is being waited-upon, maximize its priority so that it
gets executed before all other pending asyncs.
Note that we deliberately don't simply execute the async in the
calling thread in this case, to allow timed-waits to fail (which is
especially important for gimp_wait()).
(cherry picked from commit 62baffed98)
Fix indentation in gimp-parallel.{cc,h}.
Remove unused typedefs in gimp-parallel.h.
s/Gimp/Gegl/ in function-type cast in gimphistogram.c.
(cherry picked from commit 05a4437d9a)
The parallel_distribute() family of functions has been migrated to
GEGL. Remove the gimp_parallel_distribute() functions from
gimp-parallel, and replace all uses of these functions with the
corresponding gegl_parallel_distrubte() functions.
(cherry picked from commit 2736cee577)
In gimp_drawable_transform_buffer_affine(), avoid modifying the
clipping mode when transforming layer masks, since this function is
used (among other things) to transform layer masks together with
their layer, in which case they should use the same clipping mode
as the layer.
This fixes a regression introduced by commit
2ae823ba2b, causing layer masks to be
transformed with a mismatched clipping mode during layer
transforms, leading to discrepencies between the transformed layer
and the transformed mask.
This commit merely reverts the necessary part of above commit,
fixing the regression, though note that this code is really up for
some serious refactoring: the logic for determining which clipping
mode to use when is spread all over the place.
(cherry picked from commit 45fc30caa7)
Blacklist the "threaded-ml" thread, which seems to mask the
backtrace signal.
Improve signal-handler synchronozation, to avoid segfaulting when
giving up on waiting for all threads to handle the signal.
Furthermore, when one or more threads fail to handle the signal in
time, return a GimpBacktrace instance with backtraces for all the
other threads, and with empty backtraces for all the non-responding
threads, instead of returning NULL and leaking the allocated
instance. Don't blacklist threads that failed to handle the signal
in time, and instead shorten the wait period for handling the
signal, and yield execution during waiting to lower the CPU usage.
(cherry picked from commit a29d040db5)
Connect GimpImage's gimp:mask-components node to the layers node
*before* connecting the channels node, so that the image's
component mask doesn't affect the channel colors, as is the case in
2.8.
(cherry picked from commit 56920dcdbf)
... the second time you do a 180 degrees rotation
In gimp_transform_resize_adjust(), nudge the transformed layer
boundary by EPSILON toward the center, to avoid enlarging the layer
unnecessarily, as a result of numeric error amplified by rounding,
when the tranformed boundary should land on integer coordinates.
In particular, this avoids enlarging the layer when rotating by 180
degrees.
(cherry picked from commit c271992aa0)
...via hover tooltips
Use the GtkWidget::query_tooltip() signal on GimpFgBgEditor to emit an
own signal "tooltip" that has the hovered widget area as parameter.
Connect to GimpFgBgEditor::tooltip() in gimptoolbox-color-area.c and
set separate tooltips on the widget's areas, including the shortcuts
for "Swap colors" and "Default colors".
(cherry picked from commit ae9d84dd22)
Use a single segment with a "step" blending function, added in the
previous commit, instead of two separate segments, for the "FG to
BG (Hardedge)" internal gradient. This makes it simpler to change
its endpoint colors by modifying the gradient, instead of changing
the FG/BG colors.
(cherry picked from commit 84066ca26a)
... to make multi-color hard-edge gradient fills possible
Add a new "step" gradient-segment blending function, which is 0
before the midpoint, and 1 at, and after, the midpoint. This
creates a hard-edge transition between the two adjacent color stops
at the midpoint. Creating such a transition was already possible,
but required duplicating the same color at the opposing ends of two
adjacent stops, which is cumbersome.
(cherry picked from commit 68bf99e806)
Add xyY color space to the color spaces for sampling colors.
Also add code to xcf-load.c that makes sure the sample point loading
code handles unknown future GimpColorPickMode values (fall back to
PIXEL pick mode).
(cherry picked from commit 298cc57042)
Add pattern offset parameters to gimp_fill_options_create_buffer() and
pass the selection's top-left corner so that pattern fills on the same
drawable are aligned.
(cherry picked from commit 38dcb73bfc)
In the GimpBacktrace Linux backend, always use libunwind, when
available, to find symbol names, even if dladdr() or libbacktrace
had already found one. libunwind provides more descriptive names
in certain cases, and, in particular, full symbol names for C++
lambdas.
Note that, in some cases, this can result in a discrepancy between
the reported symbol name, and the corresponding source location.
(cherry picked from commit 72fc01742b)
In the GimpBacktrace Windows backend, avoid reporting meaningless
symbol addresses when failing to retrieve meaningful ones.
Unfortunately, it seems that we never get symbol addresses for
symbols that have debug information, which negatively affects the
log viewer's call graph. We're going to have to work around this.
(cherry picked from commit 52772cf3ff)
When initializing the GimpBacktrace Windows backend, set the name
of the current thread (which is assumed to be the main thread) to
the program's name, to match its name on Linux. We normally rely
on the SET_THREAD_NAME exception to set thread names on Windows,
which isn't raised for the main thread.
(cherry picked from commit 52908f397f)
In the gimp_parallel_run_async() family of functions, allow the
async callback to return without completing the async operation, in
which case the callback will be called again, until the operation
is either completed, or canceled, in which case it is aborted
(previously, returning from the callback without completing the
operation would cause it to be aborted.) It is guaranteed that all
operations of the same priority will get a chance to run, even if
some of them contuinuosly return without completing.
This allows potentially time-consuming operations to yield
execution in favor of other same-priority operations, and, in
particular, of higher-priority operations, to avoid priority
inversion. Essentially, this allows a simple form of cooperative
multitasking among async operations.
(cherry picked from commit 4969d75785)
When shutting-down gimp-parallel, cancel and/or abort any ongoing
and queued async operations, instead of finishing them (async
operations that already started executing will be canceled, but
execution will be blocked until they're finished.) This is
especially important since we're shutting down gimp-parallel before
the destruction of data factories. This commit causes any ongoing
async operations of the factories to be canceled on shutdown,
rather than waiting for them to finish normally.
(cherry picked from commit e46fdc714e)
Add a new GimpData::data_cancel() virtual function, and a
corresponding gimp_data_factory_data_cancel() function. This
function should cancel any ongoing async operations related to the
factory (i.e., included in its async set), and wait for the
operations to finish. Provide a default implementation that simply
cancels and waits on the factory's async set.
Use this function to cancel any ongoing operations during factory
destruction, and in gimp_data_factory_data_free().
Override this function in GimpFontFactory, for which we can't
really cancel font loading, and simply cancel and clear the
factory's async set without waiting for loading to finish, making
sure that nothing happens (and, in particular, that the factory
isn't being accessed, since it might be already dead) when loading
does finish.
(cherry picked from commit 6bc0b3b8ad)
In gimp_data_factory_data_foreach(), don't rely on internal
GimpData objects being sorted first (while this is currently true
for all types of GimpData, they may override the sort order.)
(cherry picked from commit 50bab438ce)
In gimp_drawable_real_{apply,replace}_buffer(), bail if the
applcation region, after intersection with the drawable and mask
extents, is empty. This avoids trying to create a GeglBuffer with
negative width/height.
(cherry picked from commit ae3c006293)
... (some sort of corruption)
In gimp_drawable_real_replace_buffer(), adjust the processed buffer
and mask_buffer regions according to the changes made to the
application region, as calculated by intersecting it with the
drawable and mask extents. This fixes wrong application position
when painting using the heal, dodge/burn, smudge, or convolve
tools, on a drawable whose origin is above/to the left of the
image's origin, and there's a selection active.
(cherry picked from commit a782acab57)
... the XCF file
Add a "saving" signal to GimpImage, which is emitted when the image
is about to be saved or exported (but before it's actually saved/
exported). Connect to this signal in tool-manager, and commit the
current tool in response (unless its GimpToolControl::preserve is
TRUE).
(cherry picked from commit ae628a8664)
When libbacktrace is available, use it to retrieve source location
information in the Linux GimpBacktrace backend.
(cherry picked from commit 7cdd1ebeef)
... and G_TYPE_INSTANCE_GET_PRIVATE()
g_type_class_add_private() and G_TYPE_INSTANCE_GET_PRIVATE() were
deprecated in GLib 2.58. Instead, use
G_DEFINE_[ABSTRACT_]TYPE_WITH_PRIVATE(), and
G_ADD_PRIVATE[_DYNAMIC](), and the implictly-defined
foo_get_instance_private() functions, all of which are available in
the GLib versions we depend on.
This commit only covers types registered using one of the
G_DEFINE_FOO() macros (i.e., most types), but not types with a
custom registration function, of which we still have a few -- GLib
currently only provides a (non-deprecated) public API for adding a
private struct using the G_DEFINE_FOO() macros.
Note that this commit was 99% auto-generated (because I'm not
*that* crazy :), so if there are any style mismatches... we'll have
to live with them for now.
Even though chosen as a parameter to gimp_image_get_xcf_version() and
not a feature within the image itself, we also want to list this reason
in the compatibility list.
(cherry picked from commit 0fa2ef9118)
Improve out-of-range check in gimp_backtrace_find_thread_by_id().
Remove unnecessary #include <exchndl.h> in gimpbacktrace-windows.c,
and revert commit 644234e99d (the
DrMingw detection happens at runtime). The Windows backend can
work without DrMingw, it just can't find all the symbols, and
doesn't provide source-location information.
(cherry picked from commit b9f1ab8f53)
The "running" attribute (readable through
gimp_backtrace_is_thread_running(), and recorded in the performance
log) specifies if the thread was in a running or suspended state at
the time the backtrace was taken. It is accurate on Linux, but
only approximated on Windows.
Adapt the performance-log-expand.py tool to maintain this attribute
(and any future thread attributes we might add).
(cherry picked from commit 78adb7c900)
The Windows backend produces full, multithreaded backtraces. When
DrMingw is available, it also provides full symbol and (where
available) source-location information. Otherwise, it provides
symbol information for most of our libraries, but not for the GIMP
binary itself.
(cherry picked from commit 667efc221d)
Add source filename and line number fields to the
GimpBacktraceAddressInfo struct, populated through
gimp_backtrace_get_address_info(). This is not currently supported
by the Linux backend, but is supported by the Windows backend,
which we'll be added in the next commit.
(cherry picked from commit a6ec857123)
This function returns information about the given address, which
is currently mostly limited to the corresponding symbol
information, but we might want to add address-specific information
in the future, such as a line number.
(cherry picked from commit 7ac87dc01e)
GimpBacktrace provides an interface for creating and traversing
multi-threaded backtraces, as well as querying symbol information.
While we already have some backtrace functionality, it relies on
external tools for the most part, and as such is rather expensive,
and is only meant for producing opaque backtraces. GimpBacktrace,
on the other hand, is meant to be relatively cheap (we're going to
use it for profiling,) and allow inspection of the backtrace data.
In the future, it might make sense to replace some, or all, of the
other backtrace functions with GimpBacktrace.
GimpBacktrace currently only supports Linux. By default, it uses
dladdr() to query symbol information, which is somewhat limited (in
particular, it doesn't work for static functions.) When libunwind
is installed, GimpBacktrace uses it to get more complete symbol
information. libunwind is currently an optional dependency, but it
might make sense to promote it to a mandatory, or opt-out,
dependency, as it's lightweight and widely available.
On other platforms, the GimpBacktrace interface can still be used,
but it always returns NULL backtraces.
(cherry picked from commit 80bf686c94)
In gimp_image_merge_layers(), explicitly fetch the graph of the top
layer's parent layer (if exists), to make sure that the top layer's
graph has a parent node. We already fetch the image graph, which
takes care of top-level layers, however, if the top layer is a
child of an invisible layer group, as is the case in the wavelet-
decompose plug-in, this is not generally enough to guarantee that
the group's graph is constructed.
(cherry picked from commit e563845174)
In GimpProjection, use an adaptive chunk size when rendering the
projection asynchronously, rather than using a fixed chunk size.
The chunk size is determined according to the number of pixels
processed during the last frame, and the time it took to process
them, aiming for some target frame-rate (currently, 15 FPS). In
other words, the chunks become bigger when processing is fast, and
smaller when processing is slow. We're currently aiming for
generally-square chunks, whose sides are powers of 2, within a
predefined range.
Note that the chunk size represents a trade off between throughput
and responsiveness: bigger chunks result in better throughput,
since each individual chunk incurs an overhead, in particular when
rendering area filters or multithreaded ops, while smaller chunks
result in better responsiveness, since the time each chunk
individual takes to render is smaller, allowing us to more
accurately meet the target frame rate. With this commit, we aim to
find a good compromise dynamically, rather than statically.
The use of adaptive chunk sizes can be disabled by defining the
environment variable GIMP_NO_ADAPTIVE_CHUNK_SIZE, in which case we
use a fixed chunk size, as before.
(cherry picked from commit a1706bbd29)
Add a boolean "chunk" parameter to
gimp_projection_chunk_render_iteration(), which determines whether
the work area should be sub-divided into chunks prior to rendering
(previously, the work area would always be sub-divided.) Only
pass TRUE when rendering the projection asynchronously, in the
render callback, and pass FALSE when rendering the projection
synchronously, in gimp_projection_finish_draw(), which is called
when flushing the projection through the GimpPickable interface.
Rendering the projection using as big chunks as possible improves
performance, while worsening responsiveness. Since responsiveness
doesn't matter when rendering synchronously, there's no reason to
render in chunks.
(cherry picked from commit 105ffc787d)
In gimp_group_layer_update_size(), never suspend drawable updates
(and, in fact, remove the option to suspend drawable updates
entirely,) and instead never update the drawable during the call to
gimp_drawable_set_buffer_full(), and flush the group's projection
*after* setting the drawable's buffer, so that any pending updates
will happen after the group's buffer and size are up-to-date.
This fixes some missed drawable updates.
(cherry picked from commit fc2c640ca2)
In GimpProjection, change gimp_projection_add_update_area() to take
coordinates in the projection's coordinate system, rather than the
image coordinate system, and move the offset adjustment to the
projectable invalidation handler.
Modify gimp_projection_projectable_structure_changed() to pass
projection-space coordinates to gimp_projection_add_update_area().
gimp_projection_get_buffer() and
gimp_projection_projectable_bounds_changed() already pass
projection-space coordinates to gimp_projection_add_update_area(),
which was wrong before, when the projection had a nontrivial
offset, but is correct now.
(cherry picked from commit 2d63bc6e0a)
Ok my previous fix was wrong (at least for the part in the macro). This
is a macro, not a function. So each time we write _reason, the call to
g_strdup_printf() is reevaluated, hence data is allocated.
The right fix is to prepend `tmp` to the list, not `_reason`.
Thanks to Massimo for the debugging, as always!
(cherry picked from commit 2912fe7c17)
ADD_REASON macro was leaking the allocated string when version_reason
return value was NULL (i.e. when we didn't care about the version
reasons).
Also we were not properly freeing all the reason strings at the end,
only the list. Use g_list_free_full() instead of g_list_free().
(cherry picked from commit 0ab682b0f5)
In gimp_projection_projectable_bounds_changed(), bail early by
calling gimp_projection_projectable_structure_changed() instead, if
the new bounds don't intersect the old bounds.
(cherry picked from commit c6b8a4213c)
In gimp_projection_projectable_bounds_changed(), which is called by
GimpProjection in response to a GimpProjectable::bounds-changed
signal, invalidate all regions of the new projection that weren't
copied from the old projection, so that they get rendered upon
flushing, instead of remaining empty.
Additionally, fix preview invalidation -- in particular, don't
directly invalidate the projectable's preview, even if preview
invalidation is already queued and chunk rendering was finished by
the boundary change, and instead always queue a preview
invalidation.
(cherry picked from commit bb5e3fd926)
Make sure we don't unnecessarily update the group layer's drawable
while flusing the group's projection during resizing, since we want
to either update the entire drawable, or avoid any updates, when
replacing the drawable's buffer. Note that explicitly supressing
updates in this case should theoretically not be necessary, but the
fact that the call to gimp_projectable_bounds_changed() can result
in reconstructing the projection (see the FIXME comment in that
function) makes it necessary in some cases nonetheless.
(cherry picked from commit bd726c96bf)
When translating group layers, there's no need to re-render the
group's projection -- we can simply update the group's offset (and
offset node) directly, and redirect any layer-stack "update"
signals to the group's drawable. This significantly improves
performance when moving groups.
(cherry picked from commit 3ff820a00a)
In GimpGroupLayer, use gimp_projectable_bounds_changed() when
updating the group layer's size, instead of reconstructing the
projection, unless reallocation of the projection has been
requested. This is more efficient, since it simply copies the
content of the projection's old buffer to the new buffer, rather
than re-rendering the graph.
(cherry picked from commit 1bb3e962f6)
In gimp_group_layer_flush(), stop any idle rendering, initiated
when a new buffer is allocated, before flushing the group's
pickable. Otherwise, the idle rendering is finished synchronously,
which unnecessarily introduces a noticeable lag.
(cherry picked from commit a4957c7c76)
... which specifies whether or not to update the drawable in
response to the buffer change.
Pass TRUE for "update" at all existing call sites, to keep the
current behavior.
(cherry picked from commit 26a8d141f6)
In GimpProjection, respond to the projectable's "bounds-changed"
signal, by reallocating the buffer, and copying the corresponding
region of the old buffer (using
gimp_tile_handler_validate_buffer_copy(), added a few commits back,
so that the relevant portion of the validate handler's dirty region
is also copied). Additionally, shift and clip all outstanding
update regions as necessary (actually, we avoid copying the buffer
when a shift is necessary, and simply reconstruct the projection;
see FIXME comment in the code.)
(cherry picked from commit fbeae36118)
... and a corresponding gimp_projectable_bounds_changed() function.
This signal can be emitted by implementers of GimpProjectable,
instead of the GimpProjectable::structure-changed signal, when the
projectable's bounds change, but its content does not -- i.e., the
old content simply gets cropped to the new bounds.
(cherry picked from commit 460c3d1349)
... which should be used to properly remove a
GimpTileHandlerValidate from a buffer, instead of using
gegl_buffer_remove_handler() directly.
Use gimp_tile_handler_validate_unassign(), instead of
gegl_buffer_remove_handler(), in gimp_projection_free_buffer().
(cherry picked from commit 12530e21b2)
In gimp_layer_convert(), avoid converting the drawable type when
the source and destination color profiles are equal, if otherwise
unnecessary. Otherwise, text layers get unnecessarily re-rendered
during conversion, and, by extension, during image duplication
(which happens when exporting to any format that requires merging
down the image). This may cause the text layer to appear
differently in the duplicated image, or even use a different font
if the original font doesn't exist.
(cherry picked from commit a826a19359)
When duplicating an image, copy the source image's is-color-managed
status to the duplicated image, instead of having the duplicated
image always be color managed. In particular, do this before
duplicating the layers, so that we don't convert the duplicated
layers from sRGB to the image's profile when duplicating an image
with a non-sRGB profile but with color management turned off.
(cherry picked from commit f38443f3b0)
In subdirs containing a generated foomarshal.h header, add the
generated sources to BUILT_SOURCES, so that they're generated
before the rest of the source files are built. Otherwise, since
there is no rule specifying the dependency between the rest of the
source files and foomarshal.h, and since foomarshal.h is not
checked into git (and hence doesn't exist when doing a clean
build), compilation of the said source files may fail if they're
built before foomarshal.h is generated.
(cherry picked from commit a5102a7dba)
Fix gimp_constrain_line() and friends to properly constrain line
angles when the image's horizontal and vertical resolutions are
different, and dot-for-dot is disabled.
(cherry picked from commit 4fefab1798)
Remember the sample point's GimpColorPickMode in the sample point
itself, so it is remembered across switching between images.
Not persistent in the XCF yet tho...
(cherry picked from commit a0129504c8)
In gimp-parallel, always flush the async-operations queue (by
executing all remaining operations on the caller thread) when
setting the async-pool thread count to 0 (as happens when setting
GEGL_THREADS=1, per the previous commit,) and not only when
shutting GIMP down. Otherwise, pending asynchronous operations
can "get lost" when setting GEGL_THREADS to 1.
Additionally, in gimp_gegl_init(), initialize gimp-parallel before
before connecting to GimpGeglConfig's "notify::num-processors"
signal, so that the number of async threads is set *before*
GEGL_THREADS, in order to avoid setting GEGL_THREADS to 1 while
async operations are still executing.
Also, allow setting the number of gimp-parallel-distribute threads
while a gimp-parallel-distribute function is running (which can
happen if gimp-parallel-distribute is used in an async operation,
as is the case for histogram calculation), by waiting for the
parallel-distribute function to finish before setting the number of
threads.
(cherry picked from commit 432a884715)
When GEGL_THREADS=1, concurrent access to the same buffer is not
safe, which can result in errors if asynchronous operations are
allowed to run in parallel to the main thread (see
https://gitlab.gnome.org/GNOME/gimp/issues/1721#note_265898.)
Disable parallel execution of asynchronous operations when
GEGL_THREADS=1 for now, to fix this. Ultimately, GEGL should be
able to remain thread-safe even when GEGL_THREADS=1. Note that we
want to execute asynchronous operations on a separate thread even
when GEGL_THREADS=1, since the goal here is mainly to avoid
blocking the main thread during their execution, rather than
speeding their execution up (in particular, it's benecifical to run
asynchronous operations in parallel even on a single-core machine,
while parallelizing GEGL operations generally isn't.)
(cherry picked from commit 408540659f)
Use gimp_babl_is_valid(), added in the previous commit, to validate
image-type/precision combinations in various functions.
(cherry picked from commit 49ca383fa4)
... (valgrind reports Invalid read)
Add gimp_babl_is_valid(), which takes a GimpImageBaseType and a
GimpPrecision, and determines whether the image-type/precision
combination is valid. Use this function to validate that loaded
XCFs use a valid type/precision combination, before trying to
create the image. Otherwise, we get a CRITICAL, and eventually a
segfault, when the combination is invalid.
Use the same function to validate the arguments of
gimp_image_new().
(cherry picked from commit a0a62656d2)
Introduce GIMP_PATTERN_MAX_SIZE (10000) and GIMP_PATTERN_MAX_NAME (256)
and validate pattern dimensions and pattern name length against them.
Add GIMP_BRUSH_MAX_NAME and validate that too.
Also make sure that the names are properly terminated, and some
cleanup.
(cherry picked from commit 9b56ca8c1d)
In gimp_color_profile_new_from_icc_profile() and
gimp_image_validate_icc_profile(), don't raise a critical when
encountering an empty profile, but rather reject it gracefully with
an error.
(cherry picked from commit 10f33b080b)
... (Invalid read reported by valgrind)
In gimp_image_parasite_validate(), don't segfault when validating
a "gimp-comment" parasite of size 0 (i.e., whose data is a 0-byte
array, not an empty string), and just consider it invalid.
(cherry picked from commit f384a0713d)
Remove the now-useless "independent" parmaeter. It is supplanted
by the new gimp_parallel_run_async_independent() function.
(cherry picked from commit 00d034a1d4)
Add an "async" field to the dashboard's "misc" group, showing the
number of async operations currently in the "running" state (i.e.,
all those GimpAsync objects for which gimp_async_finish[_full]() or
gimp_async_abort() haven't been called yet).
(cherry picked from commit aa382650a1)
Preview generation for layer groups is more expensive than for
other types of drawables, mostly since we can't currently generate
layer-group previews asynchronously. Add a preferences option for
enabling layer-group previews separately from the rest of the
layer/channel previews; both of these options are enabled by
default. This can be desirable regardless of performance
considerations, since it makes layer groups easily distinguishable
from ordinary layers.
(cherry picked from commit 30cc85fd63)
... which is an asynchronous version of
gimp_drawable_get_sub_preview().
We currently support async preview generation for drawables whose
buffer isn't backed by a GimpTileHandlerValidate tile handler
(i.e., anything other than group layers), since preview generation
fir such drawables may involve processing the corresponding graph,
which isn't thread-safe.
When the GIMP_NO_ASYNC_DRAWABLE_PREVIEWS environment variable is
defined, all drawable previews are synchronously generated.
(cherry picked from commit d79e3fbd6f)
Remove the "independent" parameter of gimp_parallel_run_async(),
and have the function always execute the passed callback in the
shared async thread-pool.
Add a new gimp_parallel_run_async_full() function, taking, in
addition to a callback and a data pointer:
- A priority value, controlling the priority of the callback in
the async thread-pool queue. 0 is the default priority (used
by gimp_parallel_run_async()), negative values have higher
priority, and positive values have lower priority.
- A destructor function for the data pointer. This function is
called to free the user data in case the async operation is
canceled before execution of the callback function begins, and
the operation is dropped from the queue and aborted without
executing the callback. Note that if the callback *is*
executed, the destructor is *not* used -- it's the callback's
responsibility to free/recycle the user data.
Add a separate gimp_parallel_run_async_independent() function,
taking the same parameters, and executing the passed callback in
an independent thread, rather than the thread pool. This function
doesn't take a priority value or a destructor (and there's no
corresponding "_full()" variant that does), since they're pointless
for independent threads.
Adapt the rest of the code to the changes.
(cherry picked from commit b74e600c12)
and remove all other tool options parent setting/unsetting and
property copying code. Also select a tool at the end of
tool_manager_init() so it is in sync with what the tool options
manager does.
(cherry picked from commit 37f69457b7)
According to some bug reports, it seems that under some (unknown)
conditions we might save an empty custom gradient file on exit (for
equally unknown reasons). The only difference in the way we save
internal data files, such as the custom gradient, compared to
gimp_data_save(), is the fact that we currently don't explicitly
close the output stream, but rather only unref it.
The output stream should be implicitly closed (and hence flushed)
upon destruction, but maybe the unreffing is not enough to
guarantee that it's actually destroyed (maybe it spawns an extra
reference for some reason, who knows.) Anyway, let's just
explicitly close it, which also gives us a chance to catch and
report any errors occursing during flushing/closing (which,
altenatively, might be the culprit).
Additionally, a few more error-reporting improvements, to match
gimp_data_save().
(cherry picked from commit a72f7f1ace)
In gimp_data_factory_finalize(), wait on the factory's async set
after canceling it, and before continuing destruction. It's not
generally safe to just abandon an async op without waiting on it
-- this is a font-specific hack, due to the fact we can't actually
cancel font loading, and GimpFontFactory is prepared to handle
this.
Instead, in gimp_font_factory_finalize(), cancel and clear the
async set, so that GimpDataFactory doesn't actually wait for
loading to finish.
In gimp_font_factory_load_async_callback(), don't try to acess the
factory when the operation is canceled, since cancelation means the
factory is already dead. On the other hand, when the opeation
isn't canceled, make sure to thaw the container even when font
loading failed, so that we always match the freeze at the begining
of the operation.
(cherry picked from commit b5890e05b8)
GIMP_BRUSH_MAX_SIZE was already defined (as 10.000 pixels per dimension,
which is big for a brush) in gimpbrush.h. Let's just use this to
validate the size returned by the header.
(cherry picked from commit b3de0bb7a5)
The flag `free_selection_string` is used to track an array of strings
with some of them being static and others allocated. This should have
been an array of boolean but we can't change it because it is public API
(though it should really not have been!).
So let's just allocate every string of the `selection` array instead,
which makes the boolean flag useless now.
We should not have essential signal connections (such as setting tool
options from brush properties) implemented in the tool options GUI
files, because they are not active until the options GUI is created.
Also, that magic is simply too hidden in the options GUI files.
Move the signal connections and the brush property copying code to
gimppaintoptions.c where is can also be done cleaner.
However, this must only be done for the main tool options instance
that is used for the GUI. Therefore, add a "gui_mode" boolean to
GimpToolOptions and set it to TRUE for all main tool options.
(this is ugly, but much less ugly and much less hidden than all the
places where code lives (like tool_manager.c) that can now be moved
into GimpToolOptions and its subclasses, and implemented cleanly
there).
(cherry picked from commit cb0e6c65d0)
Use gimp_input_data_stream_read_line_always(), instead of
g_input_data_stream_read_line(), in a bunch of places that don't
expect EOF. If we don't do that, the code assumes the GError
parameter is set by the function and returns an error indication,
causing the caller to segfault when it tries to access
error->message. Instead, we now process an empty line when EOF is
reached, which is caught by the normal parsing logic.
Additionally:
- Use gimp_ascii_strto[id]() when loading gradients, generated
brushes, and palettes, to improve error checking for invalid
numeric input.
- Improve gradient-segment endpoint consistency check.
- Allow loading palette files with 0 colors. They can be created
during the session, so we might as well successfully load them.
(cherry picked from commit 993bbd354e)
... which are similar to g_ascii_strtoll() (except that
gimp_ascii_strtoi returns a gint, and not a gint64), and
g_ascii_strtod(), however, they make error checking simpler, by
returning a boolean value, indicating whether the conversion was
successful (taking both conversion and range errors into account),
and return the actual value through a pointer.
(cherry picked from commit 3301c06163)
... which is a drop-in replacement for
g_data_input_stream_read_line(), however, it always returns a non-
NULL value when there's no error. If the end-of-file is reached,
an empty string is returned.
(cherry picked from commit e090b910c0)
gimp_layer_create_mask(): make sure we don't do a gamma conversion
when initializing the mask from a channel. This was probably not the
last place to need this fix.
Also get rid of a second switch(add_mask_type), must be some leftover
from long gone logic.
(cherry picked from commit f815a2d922)
... with a color profile other than the gimp built-in.
Remove the separate alpha-channel copy in
gimp_image_color_profile_srgb_to_pixel(). We no longer need it,
since GimpColorTransform already takes care of that itself.
We used babl_process() to copy the alpha, which would also
transform the input color from R'G'B' to the output color space.
When the same buffer was used for both input and output, this call
would overwrite the input to the subsequent
gimp_color_transform_process_pixels() call; when the output color
space was different than R'G'B', this meant we'd pass the input to
gimp_color_transform_process_pixels() in the wrong color space,
producing wrong results. This was the case when converting the
foreground color for use with the smudge tool.
(cherry picked from commit e58e2ec5dc)
...after program restart
GimpContext was always supposed to keep the names of objects (brush,
pattern, font etc.) around even if these objects don't exist, for
cases like refreshing the data in a GimpDataFactory (which worked
fine), but also for deserializing the names of objects which don't
exist *yet* (delayed loading, no-data or whatever).
This commit fixes the delayed loading case (particularly affects fonts):
gimp_context_deserialize_property(): always keep the name of the
object around when it is not found, not only in the no-data case.
gimp_context_copy_property(): always copy the object *and* its name to
the dest context.
Add GimpConfig::duplicate() and ::copy() implementations which chain
up for duplicating/copying all properties and additionally copy all
object names to the new/dest context.
(cherry picked from commit ed1e2b1524)
It seems it was simply forgotten. PROP_MASK_ALL is used at some very
central places, so this commit might fix a few subtle bugs, or
introduce new ones, everybody look for strange tool preset behavior
please :)
(cherry picked from commit b3690b48d9)
In gimp_image_set_colormap(), make sure the image's colormap always
has at least one color -- babl palette formats must have at least
one color.
If the function is called with 0 colors, use black. We still need
to support this case, in particular, since existing XCFs may have
an empty colormap, and since plug-ins can call
gimp-image-set-colormap with 0 colors.
(cherry picked from commit c16c68e63e)
... and various other palette formats
Change accidental 'G_IS_INPUT_STREAM (file)' argument validation to
'G_IS_INPUT_STREAM (input)'.
(cherry picked from commit ea6d997e56)
While most of our transform tools use an interactive transform
grid, and have similar behavior, the flip tool is an odd one out.
The new "auto straighten" function of the measure tool introduces
another tool that performs transformations, while not behaving like
the rest of the transform tools.
Factor out the parts of GimpTransformTool that handle user
interaction into GimpTransformGridTool (with corresponding
GimpTransformGridOptions, and GimpTransformGridToolUndo), and only
leave the basic transform functionality and options in
GimpTransformTool (and GimpTransformOptions).
Derive all the transform tools (and transform-tool base classes)
that previously derived from GimpTransformTool, from
GimpTransformGridTool. The one exception is GimpFlipTool, which
still derives from GimpTransformTool directly. The next commit
will derive GimpMeasureTool from GimpTransformTool as well.
...won't work with older GIMP?
Make gimp_image_get_xcf_version() return a "reason" string which lists
all reasons why the image can't be saved with compatibility for older
GIMP versions. Display the reason as tooltip on the compat hint label
in the save dialog.
(cherry picked from commit a4061a6b0d)
When creating a layer group, or a text layer, use the image's
default new-layer mode, instead of always using (non-legacy)
NORMAL.
(cherry picked from commit 7c7b6eb537)
